WiFi Based Fingerprinting Positioning Based on Seq2seq Model.

Area of Science:

• Computer Science
• Electrical Engineering
• Signal Processing

Background:

• Indoor positioning is crucial in areas lacking GPS signals.
• WiFi-based systems offer a practical solution due to widespread Access Point (AP) deployment.
• Existing methods may not fully exploit contextual data within WiFi signal sequences.

Purpose of the Study:

• To propose a novel deep learning method for WiFi fingerprinting using a seq2seq model.
• To enhance indoor positioning accuracy by utilizing sequential context information.
• To demonstrate the effectiveness of the proposed method against existing deep learning approaches.

Main Methods:

• A seq2seq deep learning model was developed for WiFi fingerprinting.
• The model learns from variable-length training sequences to capture context.
• Contextual information, such as movement patterns, is exploited for improved positioning.

Main Results:

• The proposed seq2seq model demonstrated improved performance on an open-source dataset.
• The method showed superior accuracy compared to other deep learning-based indoor positioning techniques.
• Exploiting context information within RSS fingerprints led to better localization.

Conclusions:

• The seq2seq model offers a promising advancement for WiFi-based indoor positioning.
• Leveraging context in WiFi signal sequences enhances location accuracy.
• This deep learning approach is suitable for practical indoor positioning applications.